Leandro Soriano Marcolino

I am back to posting :) But on LinkedIn (linkedin.com/in/leandro-sor…) and Mastodon (@LeandroSMarc_AI" target="_blank" rel="nofollow noopener">mastodon.social/@LeandroSMarc_…). Please follow me there :) I am also happy to follow you on those platforms.

Marcelo Lemos, my master student with UFMG, defended his work Towards a Scale-Invariant Reinforcement Learning Model for RTS. Congratulations! Thanks for the collaboration, @andertavares, @lchaimowicz, @runaldis. And thanks for @ericksonrn_, and @levilelis for joining the panel.

Another PhD success: @Kerim_AI_ defended his thesis "Synthetic Data for Machine Learning" in an intensive viva with @hrahmani60, @renatojmsdh, and @BenachourP as chair. Great work with @ericksonrn_ and @Richard11344110. He is already an assistant professor at UCA. Congratulations

My student with @yelkhatib, @mthmoog, successfully defended his PhD "Monte-Carlo Based Online planning Under Partial Observability". He will now take a position at USP, a prestigious university in Brazil. Thanks examiners and chair @lchaimowicz, @DLeslieLancs, and Ruslan Mitkov.

Happy to see that my work "Multi-agent team formation: Diversity beats strength?", about the importance of diversity in a team of voting agents, has now more than 100 citations! :) scholar.google.com/citations?view…

Double success at #AAMAS2024 :) “It Is Among Us: Identifying Adversaries in Ad-hoc Domains Using Q-valued Bayesian Estimations” and “Incentive-based MARL Approach for Commons Dilemmas in Property-based Environments” are both accepted! Final versions are coming soon.

Our paper "Reward Certification for Policy Smoothed Reinforcement Learning" was accepted at AAAI! It shows a novel technique to certify the reward of robust RL. It's one of those special cases where the paper is accepted at a top venue in the first submission. Final version soon.

My PhD student, Mário Escarce, defended his thesis "Exploring the Intersection of Human-Machine Creativity for Dynamic Generation of Arts, Music, and Games". He will take a post-doc position at UC Santa Cruz. Congratulations! Thanks @lumpenproletar, Abe Karnik, and Fabricio Goes!

Our work "Enhancing robustness in video recognition models: Sparse adversarial attacks and beyond" was accepted at Neural Networks. It presents a Bayesian-optimization approach for attacking video classifiers, which allow us to train more robust models. authors.elsevier.com/c/1iF0b3BBjKnv…

My #NeurIPS2024 paper "Information-guided Planning: An Online Approach for Partially Observable Problems" will be in the poster session tomorrow! It's available at eprints.lancs.ac.uk/id/eprint/2087…. It handles better sparse rewards, by also considering the entropy during decision-making.

Quer ganhar um livro de romance escrito por um dos nossos ex-alunos? Então participe do sorteio do livro "Vida", de autoria do Leandro Soriano Marcolino, no instagram @dcc.ufmg . #dccufmg . 🔗 linktr.ee/dccufmg

Northern dlvr.it/SzJFXQ

I've spent a really great time with all my friends and colleagues at UFMG and LU, and I'm thankful for every single second spent on this project. Special thanks to my advisors @ericksonrn_ and @LeandroSMarc_AI, and everyone I collaborated with.

I'm extremely happy to share that yesterday I successfully defended my PhD dissertation titled "Text-driven Video Acceleration: Using Visual-Textual Representations to Detect Semantics in Videos".

I am co-chairing the BMVA Synthetic Data for Machine Learning Symposium. It's a thought-provoking event that will take place on Wednesday, November 8, 2023. It is hosted at the prestigious British Computer Society (BCS) venue, located at 25 Copthall Avenue, London EC2R 7BP.

Finally updated my website with all my recent publications! :) You can now find them at: wp.lancs.ac.uk/colab/publicat…

I am co-organising the "Synthetic Data for Machine Learning" Symposium, November 8, in the British Computer Society (BCS), London. Join us for a day of discussions with great speakers about the exciting potentials of synthetic data. More information: bmva.org/meetings/23-11…

My PhD student, @Kerim_AI_, had a paper accepted at WACV, congratulations :) The paper shows a new algorithm that uses synthetic data to better stabilize (real) videos under adverse weather conditions. Final version will be available soon.

📢 BMVA Symposium: Synthetic Data for Machine Learning 🎉🚀 Join BMVA symposium in London, UK, on Wednesday, 8th November 2023! 🇬🇧 👥 Chairs: Abdulrahman Kerim - UCA, Leandro Marcolino – Lancaster University, Erickson Nascimento - Microsoft 📅 rb.gy/ty6zu 🎟️

Our work "Scale-Invariant Reinforcement Learning in Real-Time Strategy Games" was accepted at SBGames, the Brazilian conference in digital games. Great collaboration with @lchaimowicz's students and @andertavares. We were selected for the best papers track :) Final version soon.